HubsteriaNotes
Fictional demo practice · No real patient dataManage demo practice
HubsteriaNotesHelp Center

FQL Quality Lab

FQL is the Master T1 demo workflow tester. The name is retained as requested; no expansion of the acronym is assumed.

Current execution

The verified Master T1 page contains test controls for the whole patient journey or selected processes, 1–20 fictional patient personas, loaded/latest build, today's build (UTC), or a custom date (UTC). A disposable same-origin iframe exercises actual app controls and checks rendered results. Every participant gets a fresh frame/document, so in-memory fixture changes do not pollute the normal demo. Participants run sequentially, not concurrently. No real messages, claims, transactions or account creation occur.

Every requested dependent scenario creates its own patient fixture through the actual form. Fixture admission appears in the report even when only a later process was selected.

Implemented assertions:

  • Patient admission: real form save and chart/name readback.
  • Patient editing: update/name readback on the same chart ID.
  • Appointment scheduling: real dialog and named event readback in Scheduling.
  • Draft note: actual patient Create Note selector opens the editor.
  • Interview: enter all clinical sections, save draft, navigate away and reopen the same note; assert exact interview readback.
  • Sign & bill: book using the patient's saved name (including edits), open the appointment-linked note, complete and sign it, then verify exactly one Ready charge in the same patient's ledger.
  • Optional navigation sample: eight existing practice pages, nonempty content and error-screen check.
  • Optional layout containment sample: current test page at 390/768/1024/1440px, document-level overflow only.

These are bounded DOM checks, not Playwright/browser-engine certification, the Vitest suite, screenshots, visual-overlap verification, external service checks, clinical persistence or a load test. The clinical editor check certifies only fictional browser-session save/readback, not production persistence, signing-to-charge or clinical authorization. Appointment creation does not certify clinician assignment, recurrence, durations, chart synchronization or saved reload behavior. Runtime timings measure specific test actions including waits, not production Core Web Vitals.

Reports and gates

Master T1 includes an AI Workflow Planner using Lovable AI Gateway. Fictional workflow descriptions generate proposed Markdown steps, expected outcomes, evidence, edge cases, regression checks and cleanup; they are never executed automatically and have no release evidence. The endpoint validates the signed-in user and protected T1 role before calling AI. Stop propagates cancellation, partial plans remain incomplete, and credit/access failures pause requests persistently. An explicit T1 resume after access restoration clears the pause without automatically making a model request. No workflow text is stored in the database; plans can be copied or downloaded and otherwise remain page-local.

Each step records participant, process, action, expected outcome, observed outcome, path, duration and status. Failed steps produce a targeted repair prompt. Real payment/closure capabilities remain explicitly not_implemented, with implementation prompts. Dependency failures produce blocked steps. Reports include follow-up/regression prompts and can be downloaded as Markdown or copied. Last ten runs remain in the current page session only; reload clears history. Stop cancels the run and prevents approval.

Any failed, blocked, unimplemented, cancelled or empty execution returns HOLD. Passing a partial suite returns SELECTED CHECKS PASSED, never full-journey approval. No client report can merge code, publish an app or authorize production deployment. Complete production readiness remains blocked on the actual durable clinical/billing chain and a separately provisioned external browser test service.

Build identity limitation

Build identity/timestamp are stamped when the Vite configuration is evaluated, not at each live edit. They identify the loaded compilation session, not an immutable historical commit. Latest uses that loaded app. Today/date selections only run if the loaded identity matches the requested UTC date; otherwise they refuse execution. No historic snapshots or deployment registry exist. Test fixture clinical dates are separate from build date.

Verified in development

AI planner verification: a real T1 request generated a detailed intake-to-billing plan with expected outcomes and edge cases, correctly marking interviews/signing unbuilt and the plan not executed. Download, Stop before text, 390/768/1024/1920px layouts and signed-out HTTP 401 were verified. Server tests cover staff rejection before a model call, streamed T1 output, preserved 403 denial, persistent credit/access pauses and explicit resume without a model request. All 76 unit/navigation tests passed; existing whole-journey HOLD, selected-process pass and report export still worked. Final compilation passed. Gateway log listing returned no records during verification; evidence came from the real streamed response and rendered plan, not a gateway log claim.

Two-person whole journey: eight real workflow assertions passed plus navigation/layout samples; six missing clinical capabilities produced HOLD. A selected implemented-process rerun passed and exported a report. Historical date refusal was verified. Full regression suite: 59 passing unit/navigation tests at implementation time. Help center and FQL were checked at phone, tablet portrait/landscape and wide-monitor widths without document overflow. Real Master T1 session was used; no human credentials were stored in code.

Help-center front page

Public / now hosts a HubsteriaNotes-branded search/category front page inspired by the public TherapyNotes support category layout. It searches the local practice topic list and navigates to current app pages; it does not claim full-text search of all HubsteriaGuardian articles. Practice dashboard is retained at /dashboard. Original support source: https://support.therapynotes.com/hc/en-us (read October 24, 2026).

Clinical rebuild verification

A five-participant isolated whole-journey rerun on October 24, 2026 found and corrected a test-fixture identity mismatch: appointment booking had used the original name after successful demographics editing. After correction, 42 workflow/navigation/layout checks passed, including five signing-to-charge checks; five payment/closure checks remained unimplemented, so readiness remains HOLD. This exercised public fictional workflows, not authenticated real-practice or Master T1 access. A fresh public isolated FQL whole-journey run passed admission, demographics update, note editor creation, interview save/navigation/readback and appointment creation plus eight navigation and four containment samples. Signing-to-charge and real payment/closure remained not_implemented, preserving HOLD. Direct Playwright checked author signing, revision history, reviewer return/resubmit/co-sign, nine PHQ-9 answers, independent item-9 flag, clinician processing and chart readback. These checks do not establish production EHR readiness.